In an era of dramatic increase in data volume and complexity,the amount of high-dimensional data such as text,audio,and images has grown significantly,and the utilization of these data has become more frequent.Therefore,it is of great importance to design and implement an efficient high-dimensional index structure.Studies has been proved that indexes based on dimensionality reduction can improve the query efficiency of high-dimensional data.However,as the amount of data increases,it is inevitable for these techniques to face challenges such as reduced query efficiency and increased memory usage.To solve this problem,this paper proposes a high-dimensional learned index based on dimensionality reduction,which reduces the dimensionality of high-dimensional data to ordered one-dimensional data by stepwise dimensionality reduction,and then trains the learned index.Experiments on synthetic and real-life datasets show that this index structure can effectively improve query efficiency and reduce memory consumption.
关键词
机器学习/降维/学习索引
Key words
machine learning/dimensionality reduction/learned index
维普
万方数据